A Multicenter Evaluation of the Impact of Procedural and Pharmacological Interventions on Deep Learning-based Electrocardiographic Markers of Hypertrophic Cardiomyopathy.
Researchers
Journal
Modalities
Models
Abstract
Artificial intelligence-enhanced electrocardiography (AI-ECG) interpretation can identify hypertrophic cardiomyopathy (HCM) on 12-lead ECGs and is a novel way to monitor response to treatment. While initial therapies were procedural – i.e. the surgical or percutaneous reduction of the interventricular septum (SRT), medical therapy with mavacamten offers an oral alternative. To evaluate biological response to SRT, we applied an AI-ECG model for HCM detection to ECG images obtained from the same patients before and after SRT across three sites: Yale New Haven Health System (YNHHS), Cleveland Clinic Foundation (CCF), and Atlantic Health System (AHS). We also applied the model in ECGs from patients who had received mavacamten at YNHHS over the past year since its approval. A total of 70 patients underwent SRT at YNHHS, 100 patients at CCF, and 145 at AHS. Among patients at YNHHS, there was no significant change in the AI-ECG score before versus after SRT (pre-SRT: 0.55 [IQR 0.24-0.77] vs post-SRT: 0.59 [IQR 0.40-0.75], p = 0.42 by Wilcoxon signed-rank test). The AI-ECG HCM scores were higher post SRT at CCF (0.61 [IQR 0.32-0.79] vs 0.69 [IQR 0.52-0.79], p = 0.003) and AHS (0.52 [IQR 0.35- 0.69] vs 0.61 [IQR 0.49-0.70], p < 0.001). At YNHHS, among 36 patients undergoing treatment with Mavacamten, the median AI-ECG score prior to starting mavacamten was 0.46 (IQR 0.21-0.72). At the end of a median follow-up period of 237 days (IQR 140-338), the median AI-ECG score decreased significantly to 0.28 (IQR 0.11-0.50, p = 0.002). Our results demonstrate a measurable difference in the myocardial effects of SRT and mavacamten detectable on ECGs, which can reflect different mechanisms for modification of the HCM substrate. Our multicenter design also demonstrates the feasibility of an image-based AI-ECG approach for disease monitoring in HCM at the point-of-care.